Professional Documents
Culture Documents
Teacher Stress: The Mediating Role of Collective Efficacy Beliefs
Teacher Stress: The Mediating Role of Collective Efficacy Beliefs
Teacher Stress: The Mediating Role of Collective Efficacy Beliefs
Copyright !
C Taylor & Francis Group, LLC
ISSN: 0022-0671 print / 1940-067 online
DOI:10.1080/00220670903383069
T eaching is a stressful occupation (Chaplain, 2008; faction, these setbacks may be ameliorated by beliefs in the
Kyriacou, 2001), and high levels of occupational school’s collective capacity to effect change. Unfortunately,
stress have a strong effect on teachers’ perfor- validity evidence for measures of TCE and teachers’ job-
mance, career decisions, physical and mental health, and related beliefs is rare, and few studies have examined how
overall job satisfaction (Jepson & Forrest, 2006). Many beliefs in collective capabilities might influence teachers’
teachers find personal satisfaction from their work, but job job stress.
satisfaction suffers when stress from student behavior and
teaching demands are high (Manthei, Gilmore, Tuck, &
Adair, 1996). Teachers’ job stress may, however, be ame- Job Satisfaction and Job Stress
liorated by school policies, support from colleagues and Teachers’ job satisfaction—perceptions of the fulfillment
school leaders, and from a sense of collective efficacy; that derived from day-to-day work activities—is important be-
is, teachers’ perceptions that the school staff, as a group, cause it influences teachers’ performance, commitment, ab-
can effectively work together to improve student learning senteeism, physical and mental health, and overall well-
and behavior. Recent studies have shown that the col- being (Caprara et al., 2003; Judge, Thoresen, Bono, &
lective efficacy of teachers may have a positive effect on Patton, 2001). Caprara et al. labeled job satisfaction a “deci-
job satisfaction (Caprara, Barbaranelli, Borgogni, & Steca, sive element” (p. 823) that influences teacher attitudes and
2003), but studies investigating teachers’ collective effi- performance and suggested that self- and collective efficacy
cacy and job stress are rare, and evidence for validity of both contribute to teachers’ job satisfaction. Although it is
commonly used measures is lacking. There were two goals widely believed that job satisfaction is influential for teacher
in the present study: first, to test the validity and fac- and school success, surprisingly little research has examined
tor structure of multifactor measures of teachers’ collec-
tive efficacy (TCE) and job stress and, second, to examine
the relationships between TCE, job stress, and job satis- Address correspondence to Robert M. Klassen, University of Alberta,
faction for teachers at two school levels (elementary and Educational Psychology, Education North 6-102, Edmonton, Alberta
secondary). T6G 2G5, Canada. (E-mail: robert.klassen@ualberta.ca)
The Journal of Educational Research 343
the job satisfaction of teachers (Evans, 1998), with most job satisfaction. School level (i.e., elementary or secondary
studies using small samples from a limited number of schools. school) may also influence teachers’ stress and job satisfac-
Teacher stress—defined as the experience of nega- tion because organizational structure, student characteris-
tive emotions resulting from a teacher’s work (Kyriacou, tics, and academic climate differ according to the age of stu-
2001)—is inversely related to teacher self-efficacy (Betoret, dents served. For example, Wolters and Daugherty (2007)
2006; Skaalvik & Skaalvik, 2007) and positively related found elementary teachers reported higher levels of indi-
to poor teacher–pupil rapport and low levels of teacher ef- vidual efficacy for student engagement than did teachers
fectiveness (Abel & Sewell, 1999; Kokkinos, 2007). The of older children. Teacher gender may also influence per-
outcomes of teachers’ work-related stress are serious and ceptions of stress and its relationship with job satisfaction
may include burnout, depression, poor performance, absen- and collective efficacy. Recent research (e.g., Greenglass
teeism, low levels of job satisfaction, and, eventually, the & Burke, 2003; Kokkinos, 2007) has reported that female
decision to leave the profession (Betoret; Jepson & Forrest, teachers experience greater emotional exhaustion (an ele-
2006). Teacher stress is not inevitable in challenging con- ment of burnout related to job stress) than male teachers,
ditions: teachers in schools where there is good communi- possibly because of inequities in total workload (paid and un-
cation among staff and a strong sense of collegiality express paid work). Although Jepson and Forrest (2006) did not find
lower levels of stress and higher levels of commitment and gender to be related to perceived stress for teachers, Punch
job satisfaction (Kyriacou). Many studies of teacher stress and Tuettmann (1990) found that work-related stress fac-
have used single-item or unifactor measures that ask teach- tors accounted for more variance in psychological distress
ers to rate overall stress levels (i.e., “In general, how stressful for women than men. Additional research using multifactor
do you find being a teacher?”; Jepson & Forrest; Manthei measures of stress may provide new insight into the relation-
et al., 1996). However, stress from teaching is likely mul- ship of gender and teacher stress.
tifaceted. Kyriacou listed 10 sources of teacher stress, with
“maintaining discipline” and “time pressures and workload”
Present Study
(p. 29) especially relevant to overall stress. Boyle, Borg, Fal-
zon, & Baglioni (1995) noted, “workload and student mis- The first objective of the study was to test the factor struc-
behavior are the two major contributors to teacher stress” ture of the TCE, job stress, and job satisfaction measures.
(p. 62), and other researchers (e.g., Chaplain, 2008; Green- The second objective was to test a structural model of TCE,
glass & Burke, 2003) have concluded that workload and job stress, and job satisfaction, and particularly to test a
student behavior are two primary components of teacher mediation model in which it was hypothesized that TCE
stress. Using unidimensional measures of teacher stress may would mediate the effect of job stress on job satisfaction.
obscure understanding of this multifactorial construct, but This hypothesis is plausible because (a) teachers’ job stress
little research has examined the factor structure of teacher is related to job satisfaction (Manthei et al., 1996) and (b)
stress measures. teachers’ collective efficacy can be considered a job resource,
Although teachers’ job stress has been shown to directly theorized to reduce the negative influence of job stress from
influence job satisfaction (e.g., Greenglass & Burke, 2003), student misbehavior on teacher engagement, and, by exten-
TCE may act as a buffer in the relationship between job sion, job satisfaction (Bakker et al., 2007). The fit of the
stress and job satisfaction. Bakker, Hakanen, Demerouti, models was tested across genders and school levels in order
and Xanthopoulou (2007) noted that job resources such as to understand the generalizability of the model for different
social support, feedback, and collegiality boosted teacher populations and across contexts.
motivation, especially in schools where job demands such as
pupil misbehavior were high. A teacher’s confidence about
a school staff’s collective efficacy to manage student behav- Method
ior may result in lower levels of personal stress from stu- Participants
dent misbehavior. Bandura (1997) suggested that schools
are “strewn with conditions that can easily erode teachers’ Participants were 951 teachers from elementary (n =
sense of efficacy and occupational satisfaction” (p. 244), but 401) and secondary (n = 550) schools in western Canada.
that teachers in efficacious schools set higher standards for Teachers were 68% female, taught in a mix of metropoli-
pupils’ academic achievement and behavior, maintain a re- tan (55%), suburban (11%), and rural (27%) settings, and
silient sense of instructional efficacy, and spend more time reported cultural heritage as European Canadian (88%),
actively teaching and monitoring academic progress. Al- Asian Canadian (4%), First Nations (2%), and other or
though recent research has confirmed the relationship be- not listed (6%). Participants were volunteers who were ap-
tween TCE and pupil achievement (e.g., Tschannen-Moran proached at one of three large compulsory teacher conven-
& Barr, 2004), no studies to date have explored how TCE tions representing over 200 schools and were asked to fill
might influence teachers’ job stress. out a brief questionnaire on teacher motivation. Approxi-
Other factors, such as school level and teacher gender, mately 75% of teachers approached agreed to complete the
may influence the relationships among TCE, job stress, and survey.
344 The Journal of Educational Research
TABLE 1. Rotated Factor Scores from Pattern Matrix (Oblimin Rotation) for Teachers’ Collective Efficacy, Sources of
Stress, and Job Satisfaction Scale Items (N = 951)
Factor 2: Factor 3:
Factor 1: TCE for Stress from Factor 4:
TCE for student student Stress from Factor 5:
Item description instruction discipline behavior workload Job satisfaction
TCE
1. How much can teachers in your .60
school do to produce meaningful
student learning?
2. How much can your school do to get .61
students to believe they can do well
in schoolwork?
5. How much can teachers in your .72
school do to help students master
complex content?
6. How much can teachers in your .70
school do to promote deep
understanding of academic concepts?
9. How much can teachers in your .70
school do to help students think
critically?
11. How much can your school do to .56
foster student creativity?
3. To what extent can teachers in your −.57
school make expectations clear about
appropriate student behavior?
4. To what extent can school personnel −.59
in your school establish rules and
procedures that facilitate learning?
7. How well can teachers in your school −.89
respond to defiant students?
8. How much can school personnel in −.89
your school do to control disruptive
behavior?
10. How well can adults in your school −.78
get students to follow school rules?
12. How much can your school do to −.62
help students feel safe while they are
at school?
As a teacher, how great a source of stress are these factors for you. . .?
1. Difficult class .71
3. Noisy students .82
5. Maintaining class discipline .86
6. Students’ impolite behavior or .85
rudeness
2. Responsibility for student .56
achievement (e.g., exam success)
4. Too much work to do (e.g., .69
paperwork/preparing
lessons/marking)
7. Having extra duties/responsibilities .69
because of absent teachers
8. Large class size .69
Job satisfaction
1. I am satisfied with my job .88
2. I am satisfied with what I achieve at .89
work
3. I feel good at work .90
Note. Factor scores below .40 are not listed. TCE = teachers’ collective efficacy.
346
TABLE 2. Reliability Coefficients and Levels of Collective Efficacy, Job Stress, and Job Satisfaction for Male and Female Teachers in Elementary and Secondary Schools
Variable α M SD M SD M SD M SD M SD M SD M SD M SD M SD
TCE for .78 43.75 5.67 43.16 7.42 43.29 7.06 43.09 5.61 42.91 5.78 42.98 5.71 43.28 5.63 43.03 6.62 43.11 6.32
instruction
TCE for discipline .88 44.61 5.81 43.86 6.44 44.03 6.30 43.53 6.55 42.56 7.12 42.94 6.91 43.84 6.35 43.19 6.82 43.40 6.68
Stress from .67 21.85 6.43 23.65 6.36 23.25 6.41 21.49 5.92 22.98 6.34 22.40 6.22 21.60 6.07 23.30 6.36 22.76 6.31
workload
Stress from .86 21.05 6.91 22.00 6.31 21.78 6.45 20.48 6.72 22.72 6.77 21.82 6.67 20.65 6.77 22.37 6.56 21.82 6.67
standard
behavior
Job satisfaction .89 22.15 4.10 22.34 3.47 22.30 3.61 22.14 3.77 21.60 3.86 21.81 3.83 22.14 3.86 21.95 3.69 22.01 3.75
Note. TCE = teachers’ collective efficacy. TCE for discipline contained 5 items; TCE for instruction contained 6 items.
The Journal of Educational Research
The Journal of Educational Research 347
TABLE 3. Correlations for Collective Efficacy, Job Stress, and Job Satisfaction
Variable 1 2 3 4 5 6
1. Job satisfaction —
2. TCE total .47∗∗ —
3. TCE instruction .38∗∗ .90∗∗ —
4. TCE student discipline .46∗∗ .91∗∗ .63∗∗ —
5. Stress total −.27∗∗ −.12∗∗ −.08∗ −.14∗∗ —
6. Stress from workload −.21∗∗ −.07∗ −.05 −.08∗ .87∗∗ —
7. Stress from behavior −.27∗∗ −.14∗∗ −.09∗∗ −.16∗∗ .89∗∗ .54∗∗
size). Significantly more female teachers (21.3%) rated the final fit of χ 2/df = 2.91, CFI = .96, RMSEA = .05. All items
stress from workload variable at a mean score of 7 or higher, in the model showed moderate to high factor loadings.
indicating “quite a bit” or “a great deal” of stress from work- The baseline measurement models showed a good fit to
load factors. In comparison, 13.4 % of male teachers rated the data for elementary teachers, χ 2/df = 1.89, CFI = .96,
stress from workload at a mean of 7 or higher (p < .01 us- RMSEA = .05; secondary teachers, χ 2/df = 2.78, CFI = .94,
ing χ 2 comparisons). Similarly, significantly more women RMSEA = .06; male teachers, χ 2/df = 2.04, CFI = .94, RM-
(18.6%) than men (12.8%) reported feeling “quite a bit” or SEA = .06; and female teachers, χ 2/df = 2.37, CFI = .96,
“a great deal” of stress from student behavior (p < .01). RMSEA = .05. Results from the test of the baseline mod-
els confirm that the factors correspond to the hypothesized
(and EFA-generated) structure, and provide evidence of in-
Correlations ternal validity and a common factor structure of the measures
across school level and gender groups. Subsequent tests of
Table 3 presents bivariate correlations for the measures of invariance (not reported due to space constraints), showed
interest for the overall sample, with all TCE and stress scales &CFI of less than .01 when additional constraints were im-
and subscales significantly related to job satisfaction (all posed (Cheung & Rensvold, 2002), suggesting invariance
ps < .01). of pattern coefficients, structural variances, and structural
covariances across groups. Invariance testing across school
levels and genders suggests that the measures operate in a
Measurement Model similar way for these groups of teachers, and lends further
evidence of validity.
CFA was used to confirm the hypothesized factor struc-
tures of the combined variables (TCE for instruction, TCE
for student discipline, stress from workload, stress from stu- Structural Model and Mediation Analysis
dent behavior, and job satisfaction) across groups (elemen-
tary and secondary teachers; male and female teachers). The hypothesized structural model (see Figure 1) showed
Good model fit is indicated by χ 2/df < 3.0, root-mean square an adequate fit for the overall sample, χ 2(219, N = 951) =
error of approximation (RMSEA) below .08, and goodness- 644.39, χ 2/df = 2.94, CFI = .96, RMSEA = .05. The model
of-fit values above .90 (Bakker et al., 2007). Baseline models showed similar levels of fit across school level and gender
for each group were established and then compared using groups, with multigroup invariance showing no significant
increasingly stringent tests of across-group invariance. Hier- change with additional constraints (i.e., &CFI or significant
archical goodness of fit was evaluated by changes in the com- changes in χ 2/df), suggesting the structural model fit well
parative fit index (CFI), as suggested by Cheung & Rensvold across school level and gender groups.
(2002). Next, SEM tests of mediation were conducted to exam-
The initial measurement model for the overall sample ine the mediating effects of the TCE variables on the rela-
showed moderate fit, χ 2/df = 4.62, CFI = .92, RMSEA = .06. tionship between stress and satisfaction. Conducting medi-
Adding three correlated error variances (δ 1,2 , δ 3,4 , δ 7,8 ) on ation analysis with SEM has several advantages over media-
the TCE measures based on the modification indices, item- tion analysis using hierarchical regression (Shrout & Bolger,
scale characteristics (e.g., adjacent items on a scale), and on 2002). The SEM approach offers a better statistical tool for
similarities in wording among the items made a significant investigating latent variables with multiple indicators, al-
difference in the fit of the model (p for &χ 2 < .001), with a lows for the simultaneous analysis of multiple indicators, and
348 The Journal of Educational Research
FIGURE 1. Structural model of relationship among TCE, stress, and job satisfaction. Numbers in parentheses are for direct
effects (i.e., unmediated path coefficients). Probability levels listed only for path coefficients. All other loadings, variances, and
covariances significant at p < .01. Item content for TCE is available in Table 1. ∗ p < .05. ∗∗ p < .01.
controls for measurement error. The bootstrap method of mediation for TCE for student behavior. The results from
SEM-based mediation analysis was used in this study. Boot- the mediation analysis suggest teachers’ beliefs about staff
strapping involves resampling the data multiple times (2000 collective confidence to manage student misbehavior low-
times in this study) with replacement in order to generate an ers (but does not eliminate) stress resulting from student
estimation of the sampling distribution of a statistic (for fur- behavior; this finding was consistent across school level and
ther rationale for this approach, see Cheung & Lau, 2008). gender groups. One would expect that teachers’ beliefs about
The AMOS 16 program (Arbuckle, 2007) provides esti- their own capabilities to manage student behavior problems
mates of the indirect effects in mediated models, with ac- would lower job stress from student behavior; the important
companying p values, based on a two-tailed significance test. finding from this study is that teachers’ beliefs about their
A mediation effect is noted when the direct path between school staff capabilities as a group influences job stress from
stress and job satisfaction is significantly lowered with the student behavior.
introduction of the presumed mediator. TCE for instruction
did not significantly mediate the relationship between job
stress from workload and job satisfaction, with the direct Discussion
path coefficient negligibly reduced with the addition of the
The findings from the present study extend previous stud-
presumed mediator (see Figure 1 for the change in path co-
ies investigating teachers’ collective efficacy. Exploratory
efficients with the addition of the mediator). In other words,
and CFA provided support for multifactor conceptualiza-
stress from workload was not lower when teachers perceived
tions of collective efficacy and job stress, and this fac-
a higher level of school collective efficacy for instruction. In
tor structure was consistent across school level and gen-
contrast, the effect of job stress from behavior on job satis-
der groups. Job stress from workload and student behavior
faction was partially mediated by the presumed mediator of
was higher for female teachers than for male teachers, but
TCE for student behavior (p < .01). These relationships were
there was no difference in TCE, job stress, or job satisfac-
tested across groups, with similar findings of nonsignificant
tion across school levels. Finally, teachers’ beliefs in col-
mediation for TCE for instruction and significant (partial)
lective efficacy to maintain student discipline significantly
The Journal of Educational Research 349
mediated the effect of job stress from student behavior on job cially difficult conditions related to student misbehavior.
satisfaction. With student misbehavior consistently ranking as one of
Although some previous measures have conceptualized the strongest sources of teacher stress and burnout (e.g.,
TCE as a unidimensional structure (Caprara et al., 2003; Betoret, 2006; Jepson & Forrest, 2006), identifying factors
Ware & Kitsantas, 2007), results from this study support that reduce the effect of student misbehavior on teachers’
conceptualization of TCE as multidimensional, with at least job satisfaction is important because increasing job satisfac-
two distinct factors. The factor analyses of the Tschannen- tion may lead to a more motivated and stable workforce.
Moran and Barr (2004) TCE measure revealed stable fac- The results from this study suggest that support intended to
tor structure across diverse groups, with factors of TCE for assist teachers in managing student misbehavior should not
instruction and student discipline emerging from the EFA only be aimed at bolstering individual capabilities, but also at
and confirmed through the CFA. Nevertheless, the TCE building collective beliefs about managing student behavior.
measure was restricted to two domains of TCE, and other Individual teachers make a difference in student behavior,
domains such as teachers’ collective confidence to involve but the collective efforts of teachers also have a positive
leadership and support services may also reflect important influence on students (Tschannen-Moran & Barr, 2004).
collective efficacy domains. Similarly, the results from this Building the collective motivation of a school staff through
study suggest that job stress is reasonably conceptualized attention to the sources of collective efficacy—postulated as
as a multifactor construct, with distinct factors related to successful past experience, observation of successful others,
workload and student behavior. The stability of the factor verbal persuasion, and group affect (Bandura, 1997)—may
structure of these measures was shown for teachers in el- reduce the influence of teachers’ stress from student be-
ementary and secondary schools, and for male and female havior on job satisfaction, even in challenging teaching
teachers. circumstances.
The results reported in the present article show modestly
higher job stress for female teachers than for male teach-
ers, and this held true for job stress from workload and from
student behavior, and in elementary and secondary schools. Limitations
More women than men experienced “quite a bit” or “a great
deal” of stress from workload and stress from student behav- The job satisfaction variance explained by the structural
ior. These results are consistent with previous studies show- model (R2 = .34) was modest, and other variables (e.g.,
ing modest but persistent gender differences in job stress SES, teachers’ self-efficacy) likely explain additional vari-
among teachers (e.g., Antoniou, Polychroni, & Vlachakis, ance. Similarly, the direct effect of the two job stress vari-
2006; Chaplain, 2008). Several explanations may be offered ables on job satisfaction was significant but modest, and
for gender differences in stress levels. Greenglass and Burke other variables and sources of job stress that influence job
(2003) suggested that women’s elevated stress levels may satisfaction were not included in this study. The TCE for
be the result of gender differences in nonwork domains, instruction factor did not mediate the effect of job stress
with higher total workload (school plus home), and higher from workload on job satisfaction, which may have been
role conflict between work and family roles. O’Laughlin and the result of a lack of congruence in the TCE domain (i.e.,
Bischoff (2005) found that women reported less partner sup- efficacy to provide effective instruction) and the job stress
port and greater time investment in household and child domain (i.e., workload). The results from the correlational
care tasks than did men, and concluded women dispropor- data in this study should be interpreted cautiously, especially
tionately feel the impact of workload stress. These previous in terms of the directionality of the results implied by the
studies do not, however, explain female teachers’ higher lev- structural model. Although the model tested in this arti-
els of stress from student behavior. Hopf and Hatzichristou cle was based on theory postulating that collective efficacy
(1999) found female teachers to be more sensitive to exter- serves as a job resource that mediates how stress influences
nalizing behavior problems, especially from adolescent male job satisfaction, other models may also serve to explain the
students, and also found male teachers assessed children’s relationship among the variables of interest. Studies that use
interpersonal behaviors as less problematic than did female longitudinal approaches can help clarify the directionality of
teachers. These gender differences in levels of stress bear the relationships among the key variables and, in particular,
further systematic investigation. can be used to examine the causal ordering of teacher stress
Findings from the study support the hypothesis that the and collective efficacy in their effects on job satisfaction.
collective efficacy of teachers is related not only to stu- Findings from this study bear the limitations characteris-
dent achievement (e.g., Goddard, 2001), but also serves as tic of cross-sectional data gained from self-report surveys.
a job resource that mediates the effect on stress from stu- Last, the sample was restricted to Canadian teachers in one
dent behavior on job satisfaction. This finding is consistent metropolitan area and, although a large number of teachers
with the results from Bakker et al. (2007), who noted the responded from a wide variety of schools, the participants
importance of support from administrators and colleagues in this study may not represent other samples of teachers in
to help teachers cope with difficult conditions, and espe- different settings.
350 The Journal of Educational Research